Watching VinylVideo with a Raspberry Pi A+
Play back video and sound on your television using your turntable and the VinylVideo converter, as demonstrated by YouTuber Techmoan.
A brief history of VinylVideo
When demand for vinyl dipped in the early nineties, Austrian artist Gebhard Sengmüller introduced the world to his latest creation: VinylVideo. With VinylVideo you can play audio and visuals from an LP vinyl record using a standard turntable and a converter box plugged into a television set.
While the project saw some interest throughout the nineties and early noughties, in the end only 20 conversion sets were ever produced.
However, when fellow YouTuber Randy Riddle (great name) got in touch with UK-based tech enthusiast Techmoan to tell him about a VinylVideo revival device becoming available, Techmoan had no choice but to invest.
Where the Pi comes in
After getting the VinylVideo converter box to work with an old Sony CRT unit, Techmoan decided to take apart the box to better understand how it works
You’ll notice a familiar logo at the top right there. Yes, it’s using a Raspberry Pi, a model A+ to be precise, to do the video decoding and output. It makes sense in a low-volume operation — use something that’s ready-made rather than getting a custom-made board done that you probably have to buy in batches of a thousand from China.
There’s very little else inside the sturdy steel casing, but what Techmoan’s investigation shows is that the Pi is connected to a custom-made phono preamp via USB and runs software written specifically for the VinylVideo conversion and playback.
For more information on the original project, visit the extremely dated VinylVideo website. And for more on the new product, you can visit the revival converter’s website.
Be sure to subscribe to Techmoan’s YouTube channel for more videos, and see how you can support him on Patreon.
And a huge thank you to David Ferguson for the heads-up! You can watch David talk about his own Raspberry Pi project, PiBakery, on our YouTube channel.
16 comments
Draco Schrott
FYI, it has been suggested that this is a scam. TechMoan looked at the actual signal on the vinyl, and it is little more than a simplistic signal, nowhere near complex enough to encode both video and audio.
It is very likely the signal on the record is just a position signal, and the actual videos are stored on the SD card of the Raspberry Pi.
Raspberry Pi Staff Simon Long
That actually makes a lot of sense – I watched the Techmoan video and when I saw his analysis of the signal waveform in Audacity, I did wonder how what appeared to be a series of almost identical pulses could store enough data for video and audio. Given it is sold as an “art project” rather than a serious means of storing A/V, I suppose that playing back videos stored on the Pi is fair enough.
Mark
I want to say this is the craziest Pidea I’ve ever seen, yet at the same time I love it! Add a retro valve amp and an old wooden cased VHF 405 line B&W telly and you’d have a great Frankenstein-esque home entertainment system! :)
Mat
For anyone who thinks it’s fake *Watch this video* https://youtu.be/Okdh7I06jFM
..and perhaps lay off the conspiracy theories. It’s only a matter of time before you’re living in your pants in a room full of pages ripped out of the newspaper with red thread joining them all together.
Raspberry Pi Staff Simon Long
Sorry, Mat (and btw, I love your videos!), but that doesn’t prove anything at all. The suggestion was that the signal encoded on the vinyl record is a timing and identification signal – the pulses encode nothing more than a reference to a particular video frame in a particular file. If that were the case – and it seems entirely plausible – then all the tests you do above, of swapping source, varying speed, changing position etc – would all still work perfectly well.
What I would be interested to see is an analysis of the contents of a few of the individual pulses recorded on the vinyl. They look and sound like fairly simple tonebursts; unless there is some very complex data hidden in the spectrum somewhere, the idea that all they are doing is encoding a simple digital code for file and frame number seems more likely than that they are encoding real AV signals; they just didn’t sound or look complex enough for that to me.
I could be wrong – I don’t have the actual audio to analyse; just the impression I get of it from YouTube, which is far from ideal. But my gut feel is that there’s something odd going on.
Andrew B
Note: I’m not a video expert, just trying to reason through how this might work, since I was also initially surprised to see a Pi in there and wondering about potential trickery.
The fact that it degrades in quality significantly on the cassette compared to the same clip on the original record would appear to indicate that it is an actual analog recording.
You can see a zoomed in view of the signal here – https://youtu.be/UtNGVb94TFE?t=690 (best to look at it in 4K)
What I think is going on is they are taking a sawtooth wave and superimposing the video signal onto it. Each 1/2 of the sawtooth is a scan line, every time the wave crosses 0 with the sharp edge is a horizontal sync (so an edge detect can recover the hsync signal). So there are ~85 scanlines of information in each frame.
Calculate the difference between a perfect sawtooth and what is there, and you get the video signal back. In the analog world this would be done with a PLL generating the perfect wave and an analog subtraction, then an amplifier.
The sharp edge of the sawtooth wave would tend to degrade over multiple analog copies.
I am not sure how the audio information is stored.
Raspberry Pi Staff Simon Long
I hasten to point out that I’m not a video expert either!
I’m not entirely convinced that there is that serious a degradation in quality on the cassette, actually – the cassette recording is from the “Curettes” (or something like that) vinyl, whereas most of the footage on the debunking video taken from vinyl is from the Motorhead disc; looking back at the Curettes footage on the original Techmoan video, it does look to my eyes to be lower quality anyway – that said, it is very hard to judge on a few short clips that don’t necessarily match up. Also, that cassette player is a Walkman Pro – it is a very good hi-fi quality deck, so I would expect a recording made from vinyl on it to be pretty close to the original, without significant audible degradation anyway. But I’ve got no idea if it was the deck used to make the recording; if it has been well-maintained; if a good quality cassette is being used etc. Too many variables to be definitive, but I do think there is nothing in the “debunking” video which would not work perfectly well with file and frame information.
Your idea about encoding the video on a sawtooth is interesting; but that would presumably entail encoding uncompressed scan lines. In this day and age – and with the video codecs which are supported by the Pi – I think it would make far more sense (and you’d get higher quality) to use some form of digital video compression.
I honestly don’t know how this works. It is probably possible to encode some form of very low quality video into the bandwidth of an audio signal, and it would therefore be possible to store that audio signal on vinyl. But those tonebursts look like digital rather than analogue data to me, and the fact that the silences between the tones are long enough that they throw away around 10-15% of the possible bandwidth seems a strange decision; you should be fighting for every possible piece of data you can store on there, and it would be easy enough to encode the delineations between data blocks with some less wasteful method (higher amplitude pulses, for example – or if you are actually able to get valid data from the shape of a single sawtooth edge, you must be sampling fast enough that you really ought to be able to detect much shorter inter-block gaps than they have used).
Also the audio is an interesting problem; in an analogue system, they’d presumably encode it as a visible audio waveform in one channel or another so it could be easily retrieved, but there’s no sign of that (and the gaps between the tones would be silent), so my guess is that compressed digital audio is encoded somewhere on there – which again makes me suspect this is digital data.
If what had been on the disc was a continuous high frequency analogue signal, I’d have found that more credible as a low quality video and audio carrier than these tonebursts; something about them just feels wrong to me. But I might be completely wrong – as above, I am not a video expert.
I’d love to examine the contents of that SD card, or to have a closer look at that audio signal. Or just to look at the circuit on the board they have attached to the Pi – the fact it is connected via USB indicates that they are passing digital data to the Pi, which is turning it into video, and their PCB doesn’t appear to have much in the way of heavy-duty DSP or similar on it. That’s another very big “hmmm”, actually – I suspect they would need some hefty processing to decode an analogue video signal and convert it into digital frame data which the Pi could display. But reading a couple of numbers from a toneburst and passing it over USB – that’s trivial and wouldn’t require much processing at all…
Raspberry Pi Staff Simon Long
OK, so it looks as if I was wrong – someone is claiming to have decoded some of the audio on the YT video, and they say it is indeed scan lines encoded on a sawtooth waveform (as Andrew suggested), with alternate frames in the L and R channels and with the audio embedded at several times real speed in the gaps between the tone pulses (which would explain why the gaps are so big).
Still not sure exactly what they are passing to the Pi over USB; I’d guess at raw sampled data from the audio streams, with all the decode processing being done by the Pi; I’d guess their board isn’t doing much other than running a couple of ADCs to sample the incoming audio. I wonder how their earlier version worked – in 1998 it might have been easier to do the decoding all in analogue rather than with a digital system.
Brian
“someone is claiming to have decoded some of the audio on the YT video” – source, please?
Brian
Sorry Mat, there’s no way. You can’t get that much data from a vinyl running at that speed. Same goes for a Compact Cassette player running at standard speed. I also note that lifting the tone arm does not stop the signal immediately, and putting it back makes it go from black to perfect signal, no distortion. What that means is that it’s a digital signal, which, in order to contain the amount of data that is appearing on the screen, would have to be at bit rates orders of magnitude higher than a Compact Cassette or vinyl at that speed can reproduce.
What you have there are digital time code pulses referencing a position in the media on another storage medium.
P.S. the DTS digital cinema audio system, introduced in 1993 with Jurassic Park, used the same technique. It was not easy back then to encode digital audio on the slow-moving 35mm film, so they placed a time code track on the film, which synced to a CD-ROM drive that had the actual digital DTS audio.
Brian
Okay, I may be wrong, since the signal is being processed digitally on the Pi. But I’d like to see your raw signal, or get a .wav of it to download!
Andrew B
There is a Texas Instruments all-in-one USB audio chip on the PCB , so likely it’s the raw ADC data which is being passed vis USB to the Pi.
StarMariner
Some Ramblings:
My technical guess is that the Vinyl record would have to travel very fast. it doesnt look like 625 line to me , which even VCR had in the day. Laser discs where out then too in the mid 80’s just before CD. There quality was poor, so vinyl, nerrr.
In 1984 whilst I was a lad, there was a kraftwerk Tape, that could play the song and show a video, but that video was computer generated and synced with the music using a high frequency pulse that humans could just detect. Similar to PCM. Fostek 8track studios in the 1990’s used a similar technical for syncing by spliting stereo into 4 and using one of those tracks as a timing channel.
Anyway, The quality in the youtube video is to clean for the time. With todays technology, it could be done using compression and high error recovery. CDs could only store 650MB of data for video that was about 1hr of MPEG2 at SD quality , my Super 8mm film camera conversions never looked that good, then days the best for home use was a Matrox graphics card and very slow 386sx ,single not the DX with co-processor version.
VCR had tracking problems as most of my generation will tell you. So why is the field flyback generated ?
But with retro-revisited tech I think it looks good.
StarMariner
This has given me a bizzare project for anyone:
Recreate edisons wax cyclinder voice recorder, using an apple , the one you eat instead of a wax drum. I suggest something harder like a lemon or cucumber to start with.
Maybe it could then be upgraded to play apple videos. :)
Brian Tristam Williams
LaserDiscs originally spun up at 1,800 RPM – that makes a lot more sense than this dawdling vinyl platter at what, 33 1/3? 45? :-D
Kyle
I love techmoan! Love seeing my favorite things come together. Do you think the turntable can run Scratch?